Neural Networks for Machine Learning in Algorithmic Trading
| Author: | Bc. David Koubek |
|---|---|
| Year: | 2018 - summer |
| Leaders: | prof. PhDr. Ladislav Krištoufek Ph.D. |
| Consultants: | |
| Work type: | Bachelors |
| Language: | English |
| Pages: | 58 |
| Awards and prizes: | |
| Link: | https://is.cuni.cz/webapps/zzp/detail/185465/ |
| Abstract: | This thesis investigates the forecasting ability of the artificial neural network (ANN) models on five major currency pairs and compares the accuracy of several ANN architectures to the difficult to outperform random walk (RW) benchmark. The ANNs mostly stand ground against the RW, yet fail to attain significantly different results for most of the currencies in out-of-sample testing. A good predictive accuracy of a few ANN models was shown only for the Japanese yen in our results. Less complex neural network architectures supported the notion of having better generalisation capabilities for most of our datasets. |
